UTX and UTY Demonstrate Histone Demethylase-Independent Function in Mouse Embryonic Development

UTX (KDM6A) and UTY are homologous X and Y chromosome members of the Histone H3 Lysine 27 (H3K27) demethylase gene family. UTX can demethylate H3K27; however, in vitro assays suggest that human UTY has lost enzymatic activity due to sequence divergence. We produced mouse mutations in both Utx and Uty. Homozygous Utx mutant female embryos are mid-gestational lethal with defects in neural tube, yolk sac, and cardiac development. We demonstrate that mouse UTY is devoid of in vivo demethylase activity, so hemizygous XUtx− Y+ mutant male embryos should phenocopy homozygous XUtx− XUtx− females. However, XUtx− Y+ mutant male embryos develop to term; although runted, approximately 25% survive postnatally reaching adulthood. Hemizygous X+ YUty− mutant males are viable. In contrast, compound hemizygous XUtx− YUty− males phenocopy homozygous XUtx− XUtx− females. Therefore, despite divergence of UTX and UTY in catalyzing H3K27 demethylation, they maintain functional redundancy during embryonic development. Our data suggest that UTX and UTY are able to regulate gene activity through demethylase independent mechanisms. We conclude that UTX H3K27 demethylation is non-essential for embryonic viability.


Introduction
Post-translational modifications of histones establish and maintain active or repressive chromatin states throughout cell lineages. Thus, the enzymes that catalyze these modifications often have crucial roles in establishing genomic transcriptional states in developmental decision-making. Histone methylation can stimulate gene activation or repression depending on which residues are targeted. Methylation of histone H3 on Lysine 4 (H3K4me) is an active chromatin modification, while methylation on histone H3 Lysine 27 (H3K27me) is associated with repression of gene activity [1].
Utx and Uty are genetically amenable to delineate H3K27me3 demethylation dependent versus demethylation independent function in mouse development. Comparative amino acid sequence analysis of UTX and UTY reveals 88% sequence similarity in humans (83% identity) and 82% sequence similarity in mouse. Across the annotated JmjC histone demethylase domain, the similarity is at 98% and 97% for human and mouse respectively. In the TPR (tetratricopeptide repeat) domain, the similarity is at 94%. So while UTY is reported to have lost H3K27 demethylase activity, it is remarkably well conserved with respect to UTX. Recent discoveries have revealed that JMJD3 functions in macrophage lipopolysaccharide response and lymphocyte Th1 response through H3K27 demethylase independent gene regulation [36,37], suggesting that function of this family of proteins is not limited to histone demethylation. It has been hypothesized that X and Y chromosome homologs will escape X-inactivation in instances where the Y homolog has not lost functional activity and male to female dosage remains balanced [38]. Therefore, it is possible that UTX and UTY have functional overlap in H3K27 demethylase independent gene regulatory processes.
A recent publication by Lee et al. characterized heart defects in Utx homozygous embryos [39]. Cell culture experiments suggested that the phenotype resulted from H3K27 demethylase activity. Utx hemizygotes were reported to have a wide range of abnormalities, but it was not clear if any phenotypes overlap with the Utx homozygotes as no comparative data were illustrated. Given that Uty remained intact in these studies, it was not possible to conclude definitively whether Utx demethylase activity was essential for early embryonic development. Furthermore, it is not known whether mouse UTY is capable of H3K27 demethylation. The classification of UTY as having no demethylase activity is based on in vitro assays only. The possibility of in vivo demethylase activity due to other co-factors remains a possibility. Also, mouse UTY has considerable sequence divergence from human UTY. The two proteins are 75% identical overall, and 95% identical in the JmjC demethylase domain. Thus, it is possible that mouse UTY has retained demethylase activity.
In our study, we have generated mouse mutations in both Utx and Uty. Hemizygous Utx mutant male mice (X Utx2 Y + ) were runted at birth with only a small number surviving to adulthood. In contrast, Utx homozygous females (X Utx2 X Utx2 ) had severe phenotypes mid-gestation, with developmental delay, neural tube closure, yolk sac, and heart defects. Unlike homozygotes, Utx hemizygotes lack mid-gestational cardiovascular defects and are recovered in Mendelian frequencies at E18.5. Furthermore, compound hemizygous male embryos (X Utx2 Y Uty2 ) carrying mutations of both Utx and Uty phenocopy the Utx homozygotes. Thus, the disparity in hemizygous and homozygous Utx phenotypes is due to compensation by Uty in the hemizygous male embryos. We have utilized an in vivo H3K27 demethylation assay to demonstrate that mouse UTY is not capable of H3K27 demethylation. Additionally, cell culture data indicate UTX and UTY may function in gene activation as both proteins associate with the H3K4 methyl-transferase complex, the BRG1 chromatin remodeler, as well as heart transcription factors. Our results implicate a crucial H3K27 demethylase independent function for UTX and UTY in mouse embryonic development. This is the first ascribed function for UTY, and the first example of developmental redundancy for X and Y chromosome homologous genes. Notably, our data suggest the H3K27 demethylase activity of UTX is not essential for embryonic viability.

Results
Hemizygous Utx mutant male mice have reduced perinatal viability We developed mutant mouse lines to assess the contribution of UTX H3K27 demethylase function in mouse development. Two alleles for Utx were obtained from public resources. The BayGenomics gene trap line Kdm6a Gt(RRA094)Byg is designated as X UtxGT1 ( Figure 1A). RT-PCR and PCR genotyping verified the identity of this allele in both ES cells and mutant mice (Figures S1 and S2A-S2C). Additionally, we obtained the EUCOMM Kdm6a knockout line (project 26585, Kdm6a tm1a(EUCOMM)Wtsi ), designated as X UtxGT2fl , which inserts a gene trap in intron 2 along with a floxed 3 rd exon ( Figure 1A). Southern blotting and PCR genotyping verified the identity of this allele (Figures S1 and S2D-S2F). Notably, quantitative RT-PCR comparison of tail RNA from X UtxGT1 Y Uty+ versus X UtxGT2fl Y Uty+ mice demonstrated that Utx gene trap 1 is more effective than gene trap 2 (a 96% reduction compared to a 61% reduction in Figures S2C and S2F). Because X UtxGT2fl demonstrated incomplete trapping, the 3 rd exon was deleted with Cre recombinase to establish X UtxGT2D (containing both the gene trap and deleted 3 rd exon, Figure 1A). Deletion of the third Utx exon produces a frameshift and introduction of a translational stop codon when Utx is spliced from exon 2 to exon 4. X UtxGT1 and X UtxD are null alleles as UTX protein was eliminated in western blotting of these embryonic lysates ( Figure 1B, 1C). Consistent with RT-PCR data, X UtxGT2fl exhibits a reduction but not absence of UTX protein ( Figure 1D).
Heterozygous Utx female mice were crossed to wild type male mice to produce hemizygous Utx mutant males. At weaning, the hemizygous X UtxGT1 Y Uty+ , X UtxGT2D Y Uty+ , and X UtxGT2fl Y Uty+ mice all exhibited reductions of 68%, 83%, and 55% respectively from the expected genotype frequencies based on these crosses, yet expected genotype frequencies were observed at embryonic day E18.5 (Table 1). At E18.5, most of the hemizygous Utx males appeared phenotypically normal; however a small percentage of the fetuses exhibited exencephaly. At birth, the hemizygous Utx males were small and exhibited a failure to thrive phenotype. Those males that survived through this phenocritical phase reached adulthood and were fertile. Hemizygous Utx mutant males were runted compared to wild type littermates and remained smaller than controls throughout their lifespan (Figure 2A, 2B). Backcross of the Utx allele onto a C57BL/6J or 129/SvJ background affected postnatal viability, but hemizygous Utx male embryos were still readily obtained at E18.5 (Table  S1).

Author Summary
Trimethylation at Lysine 27 of histone H3 (H3K27me3) establishes a repressive chromatin state in silencing an array of crucial developmental genes. Polycomb repressive complex 2 (PRC2) catalyzes this precise posttranslational modification and is required in several critical aspects of development including Hox gene repression, gastrulation, X-chromosome inactivation, mono-allelic gene expression and imprinting, stem cell maintenance, and oncogenesis. Removal of H3K27 trimethylation has been proposed to be a mechanistic switch to activate large sets of genes in differentiating cells. Mouse Utx is an X-linked H3K27 demethylase that is essential for embryonic development. We now demonstrate that Uty, the Y-chromosome homolog of Utx, has overlapping redundancy with Utx in embryonic development. Mouse UTY has a polymorphism in the JmjC demethylase domain that renders the protein incapable of H3K27 demethylation. Therefore, the overlapping function of UTX and UTY in embryonic development is due to H3K27 demethylase independent mechanism. Moreover, the presence of UTY allows UTX-deficient mouse embryos to survive until birth. Thus, UTX H3K27 demethylation is not essential for embryonic viability. These intriguing results raise new questions on how H3K27me3 repression is removed in the early embryo. Additionally, the gene trap of X UtxGT2fl was excised with Flp recombinase to create a standard floxed exon 3 (X Utxfl ) and Cre recombination created X UtxD . (B) Western blotting of E18.5 liver demonstrates a complete loss of UTX in X UtxGT1 Y Uty+ lysates. RbBP5 was used as a loading control. (C) Western blotting of E10.5 whole embryo demonstrates a complete loss of UTX in X UtxD Y Uty+ and X UtxD X UtxD lysates. RbBP5 was used as a loading control. (D) Western blotting of E12.5 primary MEFs demonstrates a reduction of UTX in X UtxGT2fl Y Uty+ and X UtxGT2fl X UtxGT2fl lysates. RbBP5 was used as a loading control. doi:10.1371/journal.pgen.1002964.g001 Homozygous Utx females are mid-gestational embryonic lethal Human UTY lacks demethylase activity based on in vitro assays, so we hypothesized that X Utx2 X Utx2 homozygous females will phenocopy X Utx2 Y Uty+ hemizygous males in demethylase dependent function (UTX specific), but may demonstrate a more severe phenotype in demethylase independent roles. Homozygous X UtxGT1 X UtxGT1 and X UtxGT2D X UtxGT2D females were never observed at weaning or embryonic day E18.5 (Table 1), but were observed at expected genotype frequencies at E10.5. However, these embryos were dead and resorbed by E12.5 (Table 1). Notably, at E10.5 all homozygous X UtxGT1 X UtxGT1 and X UtxGT2D X UtxGT2D females were smaller in size and had open neural tubes in the midbrain region ( Figure 3A-ii, iii, vi, vii). Variation in severity of the Utx homozygous phenotypes was observed in mutant embryos, ranging from medium sized with typical E10.5 features ( Figure 3A-ii, vi) to much smaller embryos resembling the E9.5 timepoint ( Figure 3A-iii, vii). The X UtxGT1 and X UtxGT2D alleles failed to complement, as trans-heterozygous X UtxGT1 X UtxGT2D female embryos resembled individual homozygous alleles ( Figure 3A-viii). Hemizygous X UtxGT1 Y Uty+ male embryos appeared phenotypically normal at E10.5 ( Figure 3A-iv). Homozygous X UtxGT2fl X UtxGT2fl females exhibited a slight reduction in phenotypic severity; about half of the mutant embryos had open neural tubes and some survival to E12.5 (Table 1). To distinguish between embryonic and extraembryonic contribution of UTX towards the homozygous phenotype, we crossed the Sox2Cre transgene into the Utx fl background. In this cross, paternally inherited Sox2Cre expression will drive Utx deletion specifically in embryonic tissue [40]. No X Utxfl X Utxfl , Sox2Cre female embryos were recovered at E18.5, whereas X Utxfl Y Uty+ , Sox2Cre male embryos were recovered at expected frequencies (Table S2). At E10.5, X Utxfl X Utxfl , Sox2Cre embryos produced phenotypes largely identical to Utx homozygotes. In summary, Utx homozygous females demonstrate a significantly more severe embryonic phenotype in comparison to Utx hemizygous males. Mid-gestational lethality is typically associated with defective cardiovascular development. Accordingly, we observed both heart and yolk sac vasculature/hematopoietic phenotypes in Utx homozygotes. Utx homozygous mutant hearts were small and underdeveloped, and more severe embryos exhibited peri-cardial edema ( Figure 3A-ii, iii, vi, vii). The yolk sac vasculature of Utx homozygotes was pale with a reduction in the amount of vascular blood ( Figure 3B-ii). In more severe examples, homozygous yolk sacs were completely pale with an unremodeled vascular plexus ( Figure 3B-iii). Thus, abnormal cardiovascular function may be a source of lethality and developmental delay in Utx homozygous mutant embryos.

UTX and UTY have redundant function in embryonic development
The most likely explanation for the disparity between Utx hemizygotes and homozygotes is that UTY can compensate for the loss of UTX in embryonic development. We tested Utx and Uty expression in embryonic development to assess any overlap in expression patterns. Utx expression was initially gauged utilizing the B-galactosidase reporter in X Utx+ X UtxGT1 and X UtxGT1 X UtxGT1 whole mount E10.5 embryos. Utx was expressed at lower levels throughout the E10.5 embryo with a particular enrichment in the neural tube and otic placode ( Figure S3A-ii, iii, iv). In situ hybridization for both Utx and Uty demonstrated similar expression patterns characterized by widespread low-level expression with particular enrichment in the neural tube ( Figure S3B-ii, iii, v, vi). Our analysis of publicly available RNA-seq data sets [41,42] revealed similar low-levels of expression for Utx and Uty. , homozygous female E10.5 X UtxGT1 X UtxGT1 (A-ii) and X UtxGT2D X UtxGT2D (A-vi) embryos have some developmental delay including smaller size, underdeveloped hearts (white arrows), and open neural tube in the head (arrowheads). More severe embryos resemble the size and features of E9.5 embryos with cardiac abnormalities and peri-cardial edema (A-iii, vii, red arrows). Hemizygous male X UtxGT1 Y Uty+ embryos appear phenotypically normal at this stage (A-iv). The X UtxGT1 and X UtxGT2D alleles fail to complement as female X UtxGT1 X UtxGT2D embryos have identical phenotypes to homozygotes (A-viii). (B) At E10.5, homozygous X UtxGT1 X UtxGT1 female embryos exhibit either normal yolk sac vasculature with a reduction in red blood cells (B-ii) or have a completely pale yolk sac with unremodeled vascular plexus (B-iii). doi:10.1371/journal.pgen.1002964.g003 To determine whether UTY can compensate for the loss of UTX, we obtained the Welcome Trust Sanger Institute gene trap line Uty Gt(XS0378)Wtsi , designated as Y UtyGT ( Figure 4A). This line, inserted in intron 4, traps the Uty transcript in a similar position of the coding sequence as the Utx alleles (compare to Figure 1A). This gene trap line was verified by RT-PCR in ES cells and subsequent mice (Figures S1 and S2G), and it achieved a 99% reduction in Uty expression from X Utx+ Y UtyGT mouse tail RNA ( Figure 4B). Hemizygous Uty mutant males, X Utx+ Y UtyGT , were viable and fertile (Table 1). However, no compound hemizygous X UtxGT1 Y UtyGT and X UtxGT2D Y UtyGT embryos were recovered at E18.5 (Table 1). At E10.5, expected genotype frequencies of X UtxGT1 Y UtyGT and X UtxGT2D Y UtyGT males were observed, but these embryos phenocopied the developmental delay, neural tube closure, cardiac, and yolk sac defects observed in Utx homozygous embryos ( Figure 4C-iii, iv).

UTX and UTY redundancy is essential for progression of cardiac development
We performed a more detailed phenotypic assessment of Utx and Uty mutant hearts to scrutinize the extent of phenotypic overlap between X Utx2 Y Uty+ , X Utx2 X Utx2 , and X Utx2 Y Uty2 embryos. Analysis of cardiac development in similar sized E10.5 embryos ( Figure 5A-i, ii, iii, iv) revealed that Utx homozygotes and Utx/Uty compound hemizygotes failed to complete heart looping ( Figure 5A-vi, viii), whereas Utx heterozygotes and hemizygotes were phenotypically normal ( Figure 5A-v, vii). Additionally, homozygotes and compound hemizygotes had smaller hearts with a lack of constriction between the left and right ventricles. Sectioning of E10.5 hearts confirmed that Utx homozygotes and Utx/Uty compound hemizygotes have small hearts with a reduction in ventricular myocardial trabeculation and little or no initiation of interventricular septum formation ( Figure 5B-ii, iv). The outer ventricular wall of these embryos is much thinner, and the overall number of cardiomyocytes and myocardial structure is severely deficient ( Figure 5C-ii, iv). In summary, while midgestational hearts appear normal in X Utx2 Y Uty+ hemizygous males, X Utx2 X Utx2 homozygous females and X Utx2 Y Uty2 compound hemizygous males display identical deficiencies in cardiac development. Therefore, UTY compensates for the loss of UTX in hemizygous Utx mutant males, rescuing mid-gestational cardiac phenotypes. Mouse and human UTY are incapable of H3K27 demethylation in vivo UTX and UTY have redundant function in embryonic development, but it is not known whether mouse UTY is capable of H3K27 demethylation. Two independent publications demonstrated that human UTY has no catalytic activity in H3K27 demethylation in vitro [26,29]. It is possible that human UTY (and not mouse UTY) has accumulated a specific polymorphism rendering it demethylase deficient. Additionally, in vitro assays remove UTY from its natural cellular context and may lack cofactors required to promote H3K27 demethylation. Therefore, we utilized an intracellular, in vivo demethylation assay, whereby HEK293T cells transiently over-expressing the UTX carboxyterminus (encoding the JmjC and surrounding domains essential for proper structure and function) exhibit a reduction in H3K27me3 immunofluorescence levels [43]. In our assay, wild type and mutant constructs were expressed at similar levels ( Figure  S4A), and individual cells expressing similar, medium-high expression levels of each construct were selected for analysis ( Figure 6). Expression of Flag-tagged human and mouse UTX demethylated H3K27me3 and H3K27me2, while a mutation known to disrupt activity (H1146A) was unable to demethylate H3K27 ( Figure 6A and 6B, Figure S5A). Human UTX expression had no effect on other histone modifications we tested, such as H3K4me2 ( Figure S5B). In contrast, neither human nor mouse UTY were capable of demethylating H3K27me3 and H3K27me2 ( Figure 6A and Figure S5A). Cells expressing medium-to-high levels of UTY (N.100) never exhibited a reduction in H3K27me3 levels relative to nearby untransfected controls.
Our previous structural analysis of human UTX [43], combined with sequence alignments ( Figure 6C and Figure S6), suggested several amino acid substitutions in human and mouse UTY sequences might make them catalytically inactive. We introduced these mutations into the human UTX C-terminal fragment (Y1135C, T1141I, SNR1025NKS, G1172D/G1191S, I1267P, I1267V, and H1329P), and examined their effects on the in vivo demethylation activity. Of all the mutations tested, only the Y1135C and T1143I mutations completely abolished the ability of UTX to demethylate H3K27 ( Figure 6B, 6D). Complete loss of activity was similarly caused by mutations of the corresponding residues in JMJD3 (Y1377C and T1385I, Figure 6D). All qualitative data was also confirmed by immunofluorescence quantification ( Figure S4B). Y1135 is conserved throughout all H3K27 demethylases ( Figure S7), and in the crystal structure [43], it interacts with two of the three methyl groups of the H3K27me3 side chain, as well as N-oxalylglycine (NOG; an analog of the cofactor alpha-ketoglutarate) ( Figure 6E). The smaller C947 side chain of mouse UTY would not effectively maintain either interaction. T1143 is conserved throughout H3K27, H3K9, and H3K36 demethylases ( Figure S8), and also interacts with NOG ( Figure 6E). Its replacement with bulky isoleucine not only removes the hydroxyl group for interaction with alpha-ketoglutarate, but also may sterically hinder its binding. These observations are consistent with the fact that no H3K27 demethylation activity has been detected for mouse UTY, and we therefore conclude that the catalytic domain of mouse UTY has crucial amino acid replacements that render the protein incapable of H3K27 demethylation. On the other hand, we failed to identify why human UTY is catalytically inactive. Notably, restoring the 2 crucial mouse UTY polymorphisms (M-UTY C947Y, I955T) failed to recover H3K27 demethylase activity ( Figure 6B). These data suggest that unidentified structural elements in the UTY Cterminal region are also responsible for the lack of H3K27 demethylase activity.
UTX and UTY associate in common protein complexes and are capable of H3K27 demethylase independent gene regulation Although human and mouse UTY have lost the ability to demethylate H3K27, they retain considerable sequence similarity with UTX, suggesting a conserved function. To gain more insight into the overlap in UTX and UTY activities, we performed a biochemical analysis of tagged constructs to determine if UTX and UTY can associate in common protein complexes. Co-transfection of Flag tagged UTX or UTY with HA-UTX followed by immunoprecipitation demonstrates that UTX can form a multimeric complex with itself and UTY ( Figure 7A). UTX associates with a H3K4 methyl-transferase complex containing MLL3, MLL4, PTIP, ASH2L, RBBP5, PA-1, and WDR5 [23,44]. To examine incorporation into this complex, we performed immunoprecipitations with Flag tagged UTX and UTY constructs. Both UTX and UTY were capable of associating with RBBP5 ( Figure 7B). Thus, UTX and UTY are incorporated into common protein complexes.
To identify common gene targets of UTX and UTY mediated regulation we generated E10.5 mouse embryonic fibroblast (MEF) cell lines containing mutations in Utx and Uty (alleles X UtxGT2D and Y UtyGT ). The gene traps in these MEFs efficiently trapped Utx and Uty transcripts ( Figure 7C). These MEFs did not demonstrate differences in levels of global H3K27me3 ( Figure S9A). Genomewide UTX promoter occupancy has been mapped in fibroblasts [30]. Therefore, we screened our mutant MEFs for misregulated genes affected by the loss of both Utx and Uty that had been documented as direct UTX targets. The FNBP1 promoter is bound by UTX [30]. We verified UTX and UTY binding to the Fnbp1 promoter by ChIP ( Figure S9B and S9C). Fnbp1 expression was reduced to 68% of WT levels in X Utx2 Y Uty+ MEFs, but was further compromised to 42% in X Utx2 X Utx2 lines and 48% in X Utx2 Y Uty2 MEFs in which all Utx and Uty activity was lost ( Figure 7C). Analysis of E12.5 MEFs of a secondary allele (X UtxGT2fl ) also demonstrated diminished Fnbp1 expression in both  Figure S6) and include the corresponding regions in mouse UTX. Transfected cells (white arrows) over-expressing H-UTX and M-UTX (Flag immunofluorescence, green pseudocolor) exhibited global loss of H3K27me3 immunofluorescence (red pseudo-color). Cells transfected with H-UTY and M-UTY C-terminal constructs did not demethylate H3K27me3. (B) H3K27me3 demethylase assay of UTX and UTY mutant constructs. H-UTX H1146A contains a point mutation in a residue that was previously reported as defective in H3K27 demethylation. Cells expressing H-UTX H1146A had no loss of H3K27me3. Mouse UTY has a Y to C amino acid change that corresponds to position 1135 in human UTX. This UTX residue is predicted to regulate H3K27me3 binding and demethylation. Expression of H-UTX Y1135C failed to demethylate H3K27me3. Mouse UTY also has a T to I amino acid change that corresponds to position 1143 in human UTX that is predicted to regulate binding of ketoglutarate in the demethylase reaction. Expression of H-UTX T1143I failed to demethylate H3K27me3. Correction of these two altered residues in mouse Uty (M-UTY-C947Y, I955T) failed to recover H3K27me3 demethylase activity. (C) Alignment of the JmjC domains of human/mouse UTX human UTY, mouse UTY, and human/mouse JMJD3. UTY non-conservative substitutions are indicated by white boxes and residues of interest are labeled with red asterisks. The UTX mutations that were analyzed are listed above the alignment, while JMJD3 mutations are listed below the alignment. (D) HEK293T cells were transfected with C-terminal UTX and UTY constructs or full-length mouse JMJD3 constructs carrying various AA substitutions. Medium-high expressing cells (N$100 cells scored for each experiment) were scored for any visible reduction in H3K27me3 levels relative to nearby untransfected cells.  Flag vector transfection was used as a negative control for immunoprecipitation. (C) Fnbp1, a gene targeted directly by UTX, has intermediate downregulation in X Utx2 Y Uty+ MEFs (68% of WT, t-test p-value = 0.002), but was further compromised in X Utx2 X Utx2 (42% of WT, t-test p-value relative to X Utx2 Y Uty+ = 0.001) and X Utx2 Y Uty2 (48% of WT, t-test p-value relative to X Utx2 Y Uty+ = 0.02, N.4 independent MEF lines per genotype) MEFs. MEFs X Utx2 X Utx2 and X Utx2 Y Uty2 MEFs ( Figure 7D). Therefore, Fnbp1 expression is positively regulated by both UTX and UTY.
To examine the role of UTX and UTY in Fnbp1 regulation, we performed H3K27me3 ChIP on E12.5 X Utx+ Y Uty+ or X Utx2 X Utx2 MEFs ( Figure 7E). Quantitative PCR for an intergenic region served as a negative control, while HoxB1 served as a positive control for H3K27me3. Quantitative PCR demonstrated that the Fnbp1 promoter has relatively low levels of H3K27me3 with no additional accumulation in X Utx2 X Utx2 MEFs ( Figure 7E). Alternatively, H3K4me3 significantly accumulated at the Fnbp1 promoter ( Figure 7F). Notably, a loss of Fnbp1 H3K4me3 was observed in X Utx2 X Utx2 MEFs ( Figure 7F). Therefore, UTX and UTY appear to function in Fnbp1 activation by regulating promoter H3K4 methylation rather than H3K27 demethylation.

UTX and UTY can both associate with heart transcription factors to regulate downstream target genes
It has been documented that UTX can associate with heart transcription factors and with the SWI/SNF chromatin remodeler, BRG1 [39]. It has been hypothesized that UTX association with these factors mediates H3K27 demethylase dependent and demethylase independent induction of the cardiomyocyte specification program. As UTX and UTY have redundant demethylase independent function in embryonic development, we examined whether UTY can also associate with these proteins. Co-transfection of Myc-UTY with Flag-BRG1 followed by immunoprecipitation demonstrated that UTY associates with BRG1 ( Figure 8A). Myc-UTY also coimmunoprecipitated with Flag-NKX2-5, Flag-TBX5, and Flag-SRF ( Figure 8B and Figure S10A). Thus, UTY can form the same protein complexes as UTX with respect to BRG1 and heart transcription factors.
To examine function of UTY in directing activation of downstream heart transcription factor targets, we assessed the regulation of one previously characterized target, atrial natriuretic factor (ANF) [39]. Co-transfection of NKX2-5 with a ANF promoter-Luciferase reporter construct demonstrated a significant upregulation in expression off the ANF promoter ( Figure 8C). The reporter expression was significantly enhanced when NKX2-5 was co-transfected with UTY ( Figure 8C). The level of ANF reporter transcriptional enhancement was relatively weaker with UTY as compared to UTX, but this is most likely due to a reduction in the transfection efficiency of full-length UTY relative to UTX (as demonstrated in Figure 7A and 7B). UTY also significantly enhanced the ANF reporter response to TBX5 ( Figure S10B). Finally, ANF expression was significantly affected in the hearts of only X Utx2 X Utx2 and X Utx2 Y Uty2 embryos (with 52% and 57% level of expression respective to X Utx+ Y Uty+ controls, Figure 8D). X Utx2 Y Uty+ hemizygotes only had a moderate loss of ANF expression (76% expression level respective to X Utx+ Y Uty+ ) that was not statistically significant from wild type controls due to the variability in ANF expression. In summary, both UTX and UTY can associate with heart transcription factors to modulate expression of downstream targets.

Discussion
We have undertaken a rigorous genetic analysis contrasting UTX and UTY function in mouse embryonic development. In alignment with current literature, Utx homozygous females are lethal in mid-gestation with a block in cardiac development [39]. We now demonstrate that Utx hemizygous mutant males are viable at late embryonic timepoints in expected Mendelian frequencies. In fact, approximately 25% are capable of reaching adulthood. Our comprehensive phenotypic analysis of Utx hemizygous males illustrates that these embryos are phenotypically normal at mid-gestation and lack the cardiovascular dysfunction of Utx homozygous females. This stark phenotypic disparity suggests that UTY may compensate for the loss of UTX in the male embryo. Compound hemizygous Utx/Uty mutant male embryos phenocopy the cardiovascular and gross developmental delay of homozygous females, proving that UTX and UTY have redundant function in embryonic development. As we have demonstrated that mouse UTY lacks H3K27 demethylase activity in vivo, the overlap in embryonic UTX and UTY function is due to H3K27 demethylase independent activity. Given the widespread developmental delay and pleiotropy, it is difficult to assess the primary defect and tissue(s) responsible for UTX and UTY redundancy. The presence of functional UTY in Utx hemizygous males is not capable of preventing peri-natal runting and lethality, suggesting that UTX and UTY are not completely overlapping in activity. These later phenotypes could be due to H3K27 demethylase dependent activity of UTX. Furthermore, the lack of phenotype in Uty hemizygotes demonstrates the absence of any essential UTY specific function in mouse development.
The UTY Jumonji-C domain has maintained high conservation in the absence of catalytic H3K27 demethylase activity. JMJD3 mediated regulation of lymphocyte Th1 response requires an intact Jumonji-C domain, but is also not dependent on H3K27 demethylation [37]. Therefore, this domain may be an essential structural protein component, a protein binding domain, or a domain that may demethylate non-histone substrates. UTX and UTY can associate in a common protein complex and can both interact with RBBP5 of the H3K4 methyl-transferase complex. UTX, UTY, and JMJD3 all associate with H3K4 methyltransferase complexes from multiple mouse and human cell types [23,25,44,45]. The Fnbp1 promoter is bound by UTX, and gene expression is positively regulated by both UTX and UTY in MEFs. Based on our histone profiling at this locus, UTX and UTY affect the deposition of H3K4 methylation, not H3K27me3 demethylation. Therefore, the common UTX/UTY pathway in embryonic development may involve gene activation rather than removal of gene repression. JMJD3 has been linked more directly to transcriptional activation as the protein complexes with and facilitates factors involved in transcriptional elongation [46]. One cardiac target of UTX regulation, atrial natriuretic factor (ANF), was misregulated in ES cell differentiation [39]. Cell culture experiments suggest that ANF may be a target of both H3K27 demethylase dependent and demethylase independent regulation; however, this study could not distinguish UTX versus UTY were generated from the X UtxGT2D and Y UtyGT alleles. (D) Fnbp1 is similarly mis-expressed in X UtxGT2fl allelic combinations of E12.5 MEFs. X Utx2 X Utx2 and X Utx2 Y Uty2 MEFs significantly differ from X Utx2 Y Uty+ MEFs (t-test p-value = 0.05 and 0.02 respectively, N.4 independent MEF lines per genotype). (E) H3K27me3 ChIP was performed on E12.5 X Utx+ Y Uty+ control (green) and X Utx2 X Utx2 (red) MEFs. An IgG antibody control is indicated in grey. Quantitative PCR for the ChIP was performed over a negative control region (an intergenic region) as well as a positive control (HoxB1). Fnbp1 failed to accumulate H3K27me3 in X Utx2 X Utx2 MEFs (t-test p-value = 0.5, N = 4 independent MEF lines per genotype). (F) H3K4me3 ChIP was performed on E12.5 X Utx+ Y Uty+ control (green) and X Utx2 X Utx2 (red) MEFs. An IgG antibody control is indicated in grey. Quantitative PCR for the ChIP was performed over a negative control region (intergenic region) as well as a positive control (Npm1). The WT Fnbp1 promoter exhibited significant H3K4me3 accumulation, which was reduced in X Utx2 X Utx2 MEFs (t-test p-value = 0.005, N = 3 independent MEF lines per genotype). doi:10.1371/journal.pgen.1002964.g007 Anf expression was analyzed from E10.5 heart RT-PCR of various X UtxGT2D and Y UtyGT allelic combinations. A moderate, but function in ES cell differentiation. Both UTX and UTY affect the transcriptional response of an exogenous ANF reporter in the presence of heart specific transcription factors, suggesting that UTX and UTY can operate more directly by aiding in transcriptional activation of this gene rather than altering chromatin structure. Consistently, ANF expression was affected in X Utx2 X Utx2 and X Utx2 Y Uty2 embryonic hearts. UTX and UTY can both associate with the SWI/SNF chromatin remodeler BRG1, which has been hypothesized to mediate histone demethylase independent gene regulation, but the relevance and mechanism of this interaction is not known. Drosophila UTX associates with BRM (orthologous to BRG1) and CBP (a H3K27 acetyl-transferase), and the coupling of H3K27 demethylation with H3K27 acetylation may be essential for switching from a silent to active state [47].
Female cells are subject to gene silencing of one X-chromosome (X-inactivation) to balance gene dosage with males. Theory on establishing X-inactivation for X-Y chromosome homologs hypothesizes that the initial entry step is loss of function or expression of the Y homolog to create dosage imbalance [38]. This prediction also dictates that conservation of X-Y homolog function will maintain gene dosage between sexes, and the female Xhomolog will not experience pressure to inactivate. Utx and Uty represent a unique paradox to this untested theory; UTY has lost demethylation activity yet Utx escapes X-inactivation. We now demonstrate that UTX and UTY have retained embryonic redundancy, verifying the presumed correlations between Xinactivation escape and functional dosage balance. Zfx, Sox3, and Amelx represent unbalanced X-chromosome genes; they have similar hemizygous and homozygous mutant phenotypes indicating that the Y chromosome homologs have lost redundant function [48,49,50,51,52,53,54]. Zfx and Sox3 are inactivated, while the Amelx inactivation status is unknown [55,56]. Of all mouse X and Y chromosome homologs, only Utx, Kdm5c, and Eif2s3x are known to escape X-chromosome inactivation [28,56,57,58]. Interestingly, both KDM5C (SMCX) and its Y chromosome homolog, KDM5D (SMCY) have retained catalytic activity in demethylation of H3K4 di and tri-methyl residues [59,60,61,62]. In contrast to Utx X-chromosome escape driven by demethylation independent redundancy, Kdm5c may escape inactivation due to demethylation dependent redundancy. Our study is the first to demonstrate that an X-Y homologous pair that escapes X inactivation maintains functional conservation, and this escape may stem from an evolutionary benefit to maintain UTY demethylation independent function.
H3K27 demethylases are hypothesized to function in early developmental activation of ''bivalent'' PRC2 targets by coordinating H3K27 demethylation with H3K4 methylation. The H3K27 demethylation dependent phenotype (UTX specific) of Utx hemizygotes is not apparent until birth. The UTX H3K27 demethylase activity is dispensable for function in C. elegans [63]. Remarkably, the mammalian embryo, having numerous examples of H3K27me3 repression in early development, can survive to term without UTX histone demethylation. It is possible that there is further redundancy between UTX and JMJD3. JMJD3 mutant mice are not well characterized, but have been reported to be perinatal lethal with distinct features in comparison to Utx hemizygotes [32]. Therefore, it is likely that JMJD3 has distinct targets in development. Overall, the earliest H3K27 demethylation depen-dent phenotypes for all members of this gene family do not manifest until late embryonic development. This timepoint is much later than the converse early embryonic phenotypes from mutations in the H3K27 methyl-transferase complex [7,21,22]. Thus, there appears to be a lack of interplay between H3K27 methylation and demethylation in gene regulation, and the early embryonic removal of H3K27me3 from PRC2 mediated processes (such as ES cell differentiation, reactivation of the inactive Xchromosome, or establishing autosomal imprinting) may involve other mechanisms such as histone turnover or chromatin remodeling. H3K27 demethylases may certainly have crucial roles in the specification of progenitor cell populations of organ systems essential in peri-natal or postnatal viability, and genetic model systems will best assess the functional impact that H3K27 demethylation plays in these processes.

Luciferase assay
We received the ANF promoter-Luciferase reporter construct from Benoit Bruneau [69]. This construct was co-transfected in the presence of NKX2-5 or TBX5 with or without UTX or UTY. Luciferase activity was measured using the Promega Dual Luciferase Reporter Assay System on the Promega Glomax Multi Detection System. All readings were normalized to a Renilla Luciferase control that was co-transfected with all samples.

Mouse crosses
All mouse experimental procedures were approved by the University of North Carolina Institutional Animal Care and Use Committee. Utx homozygous data was generated either by crosses between Utx hemizygous males and Utx heterozygous females, or by crosses between X UtxGT2fl Y Uty+ VasaCre males and X Utx+ X UtxGT2D heterozygous females. X UtxGT2fl Y Uty+ VasaCre males were utilized because of an initial difficulty in generating X UtxGT2D Y Uty+ males and due to the efficient and specific activity of VasaCre in the male germline [70]. Utx hemizygous phenotypic data was developed from the previously mentioned homozygous crosses or through crosses between a WT male and heterozygous Utx female. Compound hemizygous Utx/Uty embryos were generated by crossing heterozygous Utx females with hemizygous Uty males. Embryos were PCR genotyped from yolk sac samples for Utx and were sexed by a PCR genotyping scheme to distinguish Utx from Uty. All primer sequences are available upon request.
Histology, in situ hybridization, and LacZ staining Histology samples, in situ hybridization, and LacZ staining were performed as described [71]. In situ hybridization probes were generated to be identical to previous literature [72]. Figure S1 Schematic of genotyping strategies for Utx and Uty alleles. (A) The X UtxGT1 allele was genotyped with a three-primer scheme spanning the insertion site in intron 3. (B) The X UtxGT2fl allele was verified by Southern blotting with an HpaI restriction digest. HpaI sites are noted as ''H'', and the 59 probe location is marked as a red box. The introduction of a novel HpaI site within the targeting cassette reduces the HpaI product from 17-Kb to 10-Kb. A three-primer scheme was designed for genotyping. Due to a deletion of intron 3 within the targeting vector, the product size of primers 1-2 will be larger in WT than in X UtxGT2fl , even with the introduction of the loxP site. Primers 3-2 will only amplify if Cre recombination takes place to delete exon 3. Table S1 Utx hemizygous genotype frequency on inbred backgrounds. Observed (Obs) and expected (Ex) frequencies of indicated genotypes (Geno) at embryonic (E) or postnatal (P) developmental stages with x 2 p-values (p-value) for the corresponding crosses to obtain each genotype. At E18.5 on the C57BL/6J background, 5 of the 8 observed X UtxGT1 Y Uty+ males were on the N8 generation. (DOC)

Table S2
Genotype frequencies of Sox2Cre driven Utx mutation. Observed (Obs) and expected (Ex) frequencies of indicated genotypes (Geno) at embryonic (E) or postnatal (P) developmental stages with x 2 p-values (p-value) for the corresponding crosses to obtain each genotype. (DOC)